Introduction
Actio (AI Context Toolkit) is an AI-sidecar framework that gives AI coding agents (Cursor, Copilot, etc.) structured context so they follow your architecture and rules instead of guessing.
The problem
AI agents hallucinate or drift because:
- Context is unstructured — they don't know which files matter.
- Architecture isn't explicit — rules live in heads or scattered docs.
- There's no single entry point — agents scan randomly.
The solution
Actio adds a small sidecar to your repo: an actio/ directory and an ENTRYPOINT.yaml entry file. The sidecar is the source of truth for agents.
- Deterministic entry point: Agents read
actio/router.yamlfirst. - Structured routing: The index points to architecture, interfaces, patterns, rules, and task guides.
- Validation: The CLI checks that the sidecar is complete and consistent.
What you get
| Feature | Description |
|---|---|
| Sidecar layout | actio/ with architecture/, interfaces/, patterns/, rules/, tasks/ |
| Index router | router.yaml routes agents to the right docs per domain and task |
| CLI | actio create, actio init, actio validate, actio doctor, actio mcp |
| Schema validation | Index and referenced files are validated |
| Plugins | YAML plugins under actio/plugins/ for extra checks |
| MCP server | Expose Actio context to AI tools over stdio |
Who it's for
- Teams using Cursor, GitHub Copilot, or other AI coding tools.
- Repos that need consistent architecture and clear boundaries.
- Anyone who wants less hallucination and more predictable AI-assisted code.